Loading…
DN

Dat Nguyen

McGill University
Evaluating predictive models of invasive species in changing environments

Species distribution models (SDMs) are widely used numerical tools that statistically relate species occurrences to its environment to predict distributions across landscapes. Despite the regular use of SDMs in ecology, recent publications have suggested that extrapolation to novel environmental conditions may fail, either spatially or temporally. For my project, I conduct an extensive assessment of the predictive ability of SDMs using invasive species across disjoint distributions, and examine possible predictors of model failure. Using global occurrence records for hundreds of invasive species of various taxonomic groups, models are fit to their native ranges, and the predictive ability of these models are then examined when extrapolating to spatially disjoint invasive ranges. Predictors of performance loss are examined using linear mixed models, including factors associated with the database and models, taxonomy, propagule pressure, environmental measures and spatial autocorrelation. My project provides a more general understanding of the reliability of SDMs and where they may fail, which will be important in the future development of predictive models in cases of global change and extrapolation to new locations.